Deconstructing Agile

Every specialization has its own jargon, and IT is no different—but many times it seems that techies love to co-opt regular English words and give them new meanings. Not only does this practice lead to confusion in conversations with non-techies, but even the techies often lose sight of the difference between their geek-context definition and the real world definition that “normal” people use.

In our Licensed ZapThink Architect course, for example, we spend far too long defining Service. This word has far too many meanings, even in the world of IT—and most of them have little to do with what the rest of the world means by the term. Even words like business have gone through the techie redefinition process (in techie-speak, business means everything that’s not IT).

It comes as no surprise, therefore, that techies have hijacked the word Agile. In common parlance, someone or something is agile if it’s flexible and nimble, especially in the face of unexpected forces of change. But in the world of technology, Agile (Agile-with-a-capital-A) refers to a specific category of software development methodology. This definition dates to 2001 and the establishment of the Agile Manifesto, a set of general principles for building better software. In the intervening decade, however, Agile has taken on a life of its own, as Scrum, Extreme Programming, and other Agile methodologies have found their way into the fabric of IT.

Such methodologies indubitably have strengths, to be sure—but what we have lost in the fray is a sense of what is particularly agile about Agile. This point is more than simple semantics. What’s missing is the fundamental connection to agility that drove the Manifesto in the first place. Reestablishing this connection, especially in the light of new thinking on business agility, is essential to rethinking how IT meets the ever-changing requirements of the business.

The Paradox of Business Requirements

How do techies know what to build? Simple: ask the stakeholders (the “business”) what they want. Make sure to write down all their requirements in the proverbial requirements document. Now build something that does what that document says. After you’re done, get your testers to verify that what you’ve built is what the business wanted.

Or what they used to want.

Or what they said they wanted.

Or perhaps what they thought they said they wanted.

And therein lies the rub. The expectation that the business can completely, accurately, and definitively describe what they want in sufficient detail so that the techies can build it precisely to spec is ludicrously unrealistic, even though such a myth is inexplicably persistent in many enterprise IT shops to this day. In fact, the myth of complete, well-defined requirements is at the heart of what we call the “waterfall” methodology, illustrated in the figure below.

 Waterfall Methodology Slide

In reality, it is far more common for requirements to be poorly communicated, poorly understood, or both. Or even if they’re properly communicated, they change before the project is complete. Or most aggravating at all, the stakeholder looks at what the techies have built and says, “yes, that’s exactly what I asked for, but now that I see it, I realize I want something different after all.”

Of course, such challenges are nothing new; they gave rise to the family of iterative methodologies a generation ago, including the Spiral methodology, IBM’s Rational Unified Process, and all of the Agile methodologies. By taking an iterative approach that involves the business in a more proactive way, the reasoning goes, you lower the risk of poorly communicated, poorly understood, or changing business requirements. The figure below illustrates such a project.

 Rational Unified Process Slide

In the above diagram the looped arrows represent iterations, where each iteration reevaluates and reincorporates the original requirements with any further input the business wants to contribute. But even with the most agile of Agile development teams, the process of building software still falls short. It doesn’t seem to matter how expert the coders, how precise the stakeholders, or how perfect the development methodology are, the gap between what the business really needs and what the software actually does is still far wider than it should be. And while many business stakeholders have become inured to poorly fitting software, far more are becoming fed up with the entire situation. Enough is enough. How do we get what we really want and need from IT?

Changing the Way We Deal with Change

Even the most Agile development teams still struggle with the problem of changing requirements. If requirements evolve somewhat during the course of a project, then a well-oiled Agile team can generally go with the flow and adjust their deliverables accordingly, but one way or the other, all successful software projects come to an end. And once the techies have deployed the software, they’re done.

Have a new requirement? Fund a separate project. We’ll start over and include your new requirements in the next version of the project we already finished, unless it makes more sense to build something completely new. Sometimes techies can tweak existing capabilities to meet new requirements quickly and simply, but more often than not, rolling out new versions of existing software is a laborious, time-consuming, and risky process. If the software is commercial off the shelf (COTS), the problem is even worse, since the vendor must base new updates on requirements from many existing customers, as well as their guesses about what new customers will want in the future. The figure below illustrates this problem, where the software project represented by the box can be as Agile as can be, and yet the business still doesn’t get the agility it craves. It seems that Agile may not be so agile after all.

 Agile Software Development Slide

The solution to this problem is for the business to specify its requirements in a fundamentally different way. Instead of thinking about what it wants the software to do, the business should specify how agile it expects the software to be. In other words, don’t ask for software that does A, B, C or whatever. Instead, tell your techies to build you something agile.

We call this requirement the metarequirement of agility—a metarequirement because agility applies to other requirements: “build me something that responds to changing requirements” instead of “build me something that does A, B, and C.” If we can build software that satisfies this metarequirement, then our diagram looks quite different:

 metarequirement of agility

Because the software in the above diagram is truly agile, it is possible to meet new requirements without having to change the software. Whether the process inside the box is Agile is beside the point. Yes, perhaps taking an Agile approach is a good idea, but it doesn’t guarantee the resulting software is agile.

The ZapThink Take

Sounds promising, to be sure, but the devil is in the details. After all, if it were easy to build software that responded to changing requirements, then everybody would be doing it. But there’s a catch. Even if we built software that could potentially meet changing requirements, that doesn’t mean that it actually would—because meeting changing requirements is part of how you would use the software, rather than part of how you build it. In other words, the users of the software must actually be part of the agile system. The box in the diagram above doesn’t just represent software anymore. It represents a system consisting of software and people.

Such software/people systems of systems are a long-standing fixture in ZapThink thinking. In fact, ZapThink frequently talks about agility in the context of SOA. With SOA, IT publishes business Services that represent IT capabilities and information, and the business drives the consumption and composition of those Services. In mature SOA deployments, policies drive the behavior of Services and their compositions. If you want to change the behavior, change the policy. In other words, SOA is governance-driven, and governance applies to the behavior of both people and technology.

Agile architectural approaches like SOA, therefore, focus on implementing governance-driven technology/people systems that support changing requirements over time. The challenge, of course, is actually building such systems that meet the business agility metarequirement. Where in this system do we put the agility? It’s not in any part of the system. Instead, it’s a property of the system as a whole—what we call an emergent property. If you’ve been following ZapThink, you’ve heard this story before: business agility is an emergent property of the combination technology/human system we call the enterprise.

In other words, we started by deconstructing the notion of Agile and ended up with Enterprise Architecture, because what is Enterprise Architecture but best practices for designing and building the enterprise to better meet changing requirements over time? This is not the static, framework-centric EA from years past that presuppose a final, ideal state for the enterprise. We’re talking about a new way of thinking about what it means to architect technology-rich organizations to be inherently agile.

Want to learn more? Stay tuned for our new book The Agile Architecture Revolution: How Cloud Computing, REST-Based SOA, and Mobile Computing are Changing Enterprise IT, to be published by John Wiley & Sons in the spring of 2013.

Photo by Rhys Asplundh.